Odometer Machine Learning Algorithm

[seanlowe]

We need to develop an algorithm where, given a list of "Odometer Candidates", picks the best one or the one that is most likely the odometer.

Some things to filter candidates are:

1. Value > 1000 (some instrument clusters have x1000 on it next to rpm)
2. No decimals (.)
3. No colons (:)
4. No letters with digits

.. and any other restrictions you can think of that make it better.

[seanlowe]

Algorithm requirements:

• first step would be to go through all candidates and remove any letters from them
  • if it encounters any '.' or ':' in there, disregard that candidate entirely (as it's probably a trip mileage or a clock)
• after all candidates have been filtered, if any candidates are now just empty strings, disregard those.
• then it would disregard any that are lower than 1000
• then the number of options would be relatively small and would be able to guess the best option